The paper is under review, and the complete code will be released later

LeadDisFlow

Discovery of EP4 antagonists through image-guided deep learning workflow  

Create conda env

conda env create -f environment.yaml

Prepare datasets for AnDisFlow

Download datasets from "Data availability".

 

First Step --> AnDisFlow-G

🌱Train the AnDisFlow-G

Slice the EP4 datasets for train

python slice_data2fragmentes.py --input_path your_ligands_file.csv --output_path your_ligands_file_fragments.csv

Construct sliced datasets to muti-smi files

python fragments2mutiSmi.py --input_path your_ligands_file_fragments.csv --output_path datasets/

Traning empty model from one sliced smi files

python create_emptyModel.py ---input_path datasets/one.smi --output_path datasets/model.empty

Begin training

python train_model.py --input_path datasets/model.empty --output_path datasets/model.trained -s datasets/train_fragments/

🪴Inference from the AnDisFlow-G

Slice the EP4 datasets for inference

python slice_data2fragmentes.py --input_path your_inference_file.csv --output_path your_inference_file_fragments.csv

Begin inference

python sample_scaffolds.py -m data_EP4/EP4_decorator/models/model.trained.90 -i data_EP4/EP4_decorator/CN109836434B_slice.smi -o data_EP4/generated_and_screen/CN109836434B_generated.csv -r 64 -n 64 -d multi --of csv 

 

Second Step --> AnDisFlow-S

🍏Train the AnDisFlow-S

cd Screen

Use the data for screen in folder(DataForScreen), and download pre-trained model and push it into the folder ckpts/ link: https://drive.google.com/file/d/1wQfby8JIhgo3DxPvFeHXPc14wS-b4KB5/view

Finetune:

python finetune.py --gpu ${gpu_no} \
                   --save_finetune_ckpt ${save_finetune_ckpt} \
                   --log_dir ${log_dir} \
                   --dataroot ${dataroot} \
                   --dataset ${dataset} \
                   --task_type ${task_type} \
                   --resume ${resume} \
                   --image_aug \
                   --lr ${lr} \
                   --batch ${batch} \
                   --epochs ${epoch}

More info about Imagemol, you can reference: https://github.com/HongxinXiang/ImageMol